使用混合反應時間模型在加入虛題的問卷中以識別作答類型

Accounting for Careless Responding in Survey Questionnaires with a Bogus Item and Response Time

研究人員在問卷蒐集後的資料分析中，未認真作答的受試者（意即可能不看題目就作答）之問卷會影響到參數估計的有效性，進而產生不正確的推論。問卷中設置虛題在許多研究中很普遍，往往可以有效地辨別未認真作答的受試者。而試題反應時間是另一個常見的工具，提供受試者的應試行為。本研究的目標是如何同時將試題反應時間和虛題設計結合而更有效地辨別出未認真作答的受試者，研究者提出混合反應時間模型，將反應時間放在模型中的潛在類別變項，而非一般傳統作法將反應時間視為控制變量。本研究不只將受試者分為認真及不認真作答，而是更有效地分出第三類受試者：答對虛題但事實上是屬於不認真作答的類別。最後，研究者以提出的模型來分析一筆實徵資料（答錯虛題的比例高達22％）。

A valid inference for a survey is always based on the efficient estimation, which is often threatened and distorted by careless responses. A bogus item is commonly used in the survey questionnaire to detect the careless respondents, who answer without reading the items. The response time is another usual method to identify the careless responses and its effect in enhancing item parameter estimation has been verified in the educational testing. To reap the benefits of both the bogus item and the response time, we propose a mixture response time model to analyze the survey questionnaire including a bogus item. The proposed inference procedure not only distinguishes the attentive and the careless respondents but also effectively discovers the third responding behavior, those who answer the bogus item correctly but indeed belong to the careless group. Simulation studies show that the standard errors of the item parameters decrease for the samples of sizes being 500 or greater, together with the response time. We illustrate our proposed method with an analysis of a large proportion of careless responding behavior (22%) in a real data set.